RSNA 2014 3D Printing (Hands-on) Training Guide

RSNA 2014 3D Printing (Hands-on) Training Guide
RSNA 2014
3D Printing (Hands-on)
Training Guide
Frank J. Rybicki
Course Director
Applied Imaging Science Lab
Boston, MA, USA
Course Faculty
Tianrun Cai
Andreas Giannopoulos
Gerald Grant
Amir Imanzadeh
Tatiana Kelil
Hansol Kim
Peter C. Liacouras
Dimitrios Mitsouras
Tim Mueller
Catherine H. Phillips
Beth A. Ripley
Asha Sarma
Nicole Wake
1
Introduction
Amir Imanzadeh and Frank J. Rybicki
The term “3D printing” refers to fabrication of a tangible object from a digital file by a 3D
printer. Materials are deposited layer-by-layer and then fused to form the final object. Many
different materials can be used to fabricate the model. Additive Manufacturing (AM), Rapid
Prototyping (RP), and Additive Fabrication (AF) are synonyms for 3D printing. According to
the most recent classification by American Society of Testing and Materials (ASTM), there
are seven major types of 3D printing technology. Although these technologies share
similarities, they differ in speed, cost, and resolution of the product.
Digital Imaging and Communications in Medicine (DICOM) image files cannot be used
directly for 3D printing; further steps are necessary to make them readable by 3D printers.
The purpose of this hands-on course is to convert a set of DICOM files into a 3D printed
model through a series of simple steps. Some of the initial post-processing steps may be
familiar to the radiologist, as they share common features with 3D visualization tools that are
used for image post-processing tasks such as 3D volume rendering.
Most 3D printed models are derived from DICOM images generated from CT scans. Images
should be reconstructed from isotropic voxels with slice thickness less than or equal to 1.25
mm. For 3D printing, image post-processing has both similarities to and substantial
differences from methods used by radiologists for 3D visualization.
 As in 3D visualization, specific software packages enable segmentation of
DICOM images using semi-automated and manual segmentation algorithms,
allowing the user to demarcate desired parts. The most commonly used tools are
thresholding, region growing, and manual sculpting.
 The segmented data are then exported in a file format that can be recognized by
3D printers. In essence, this process is conversion of 2D images to 3D triangular
facets that compose a mesh surface. To date, the most widely used format is
Standard Tessellation Language (denoted by the file extension “STL”).
 In most cases, the STL output is not optimized for printing and further refinement
is required. This refining step may be unfamiliar even to radiologists versed in 3D
visualization; Computer Aided Design (CAD) software is used to perform steps
such as “wrapping” and “smoothing” to make the model more homogeneous.
 A key part of 3D printing is choosing the appropriate hardware technology and
material. There are several considerations in choosing which technology to use,
such as availability, cost, speed, biocompatibility, and most importantly
anticipated usage of the product (e.g., a model for surgical planning versus a
custom made implant).
Our ultimate goal is to educate participants about the capabilities of 3D printing and, through
this hands-on-exercise, provide an initial working knowledge of how it is performed.
This session focuses on image post-processing of DICOM image files generated from a CT
scan for 3D printing. Participants will learn to segment simple to moderately complicated
structures and prepare them for 3D printing. Using this handout as a guide, we will teach
1
participants to use three software packages, Mimics and 3-matic (Materialise) and Objet
Studio (Stratasys). The following descriptions are provided by the vendors *.
Mimics is an image-processing package that interfaces between 2D image data (e.g., CT,
MRI) and 3D engineering applications. Mimics is widely used in academics, hospitals, and
industry for 3D printing as well as for anatomical measurements, 3D analysis, Finite Element
Analysis, patient-specific implant or device design, and surgical planning or simulation.
Within Mimics, users can segment any region of interest that can be seen in the medical data
and accurately create a 3D model of patient anatomy.
3-matic is a Computer Aided Design (CAD) package dedicated for use with anatomical data.
It can perform common CAD operations directly on triangulated STL files. It can also be used
to optimize the triangle mesh so the anatomical models can be used in a finite element
package.
Objet Studio makes building high-quality, highly detailed and accurate models a fast,
efficient operation. Objet Studio supports STL and SLC files from any 3D CAD application.
The software offers simple “click & build” preparation and print tray editing. It provides easy,
accurate job estimation and full job control, including queue management. The software also
features powerful wizards that facilitate and speed system maintenance.
*
Please note that the course director and staff do not endorse or receive funding or royalties
from 3D printing products. Similarly, the RSNA does not endorse specific commercial
products. However, the company trade names are inherently part of this course because using
them is the only way to effectively communicate the educational message.
2
We will segment the abdominal aorta and prepare an STL file that can be printed. Because
time in this Hands-On Session is limited, the RSNA computers have the CT DICOM images
pre-loaded, and the software has already been launched. In practice, these simple initial steps
require additional understanding of the software (described in APPENDIX I).
Before we begin Task A, we will introduce the Mimics software environment, specifically the
menus, toolbars, windows, and shortcuts.
Shortcut
Action
Scroll wheel (center mouse)
OR Shift+right click+drag
Ctrl+right click+drag
Pan: Move the mouse while keeping the center click
pressed
Zoom: Move the mouse vertically while keeping the
buttons pressed to zoom in and out
Go to next slice
Go to previous slice
ArrowUp/Scroll wheel up
ArrowDown/Scroll wheel
down
PageUp
PageDown
CTRL + L
SPACE
Backspace
Right click + drag on
images
Skip 10 slices upward
Skip 10 slices downward
Make slice indicators visible/invisible
Zoom window with cursor to full screen
Switch between two window states
Adjusts contrast window in images
3
Task A. Creating a Mask of (Segmenting) the Aorta
WHAT YOU ARE DOING: Segmenting the aorta. The term “segmentation” describes the task of
identifying specific voxels in a region of interest such as the aorta. You will isolate the contrastenhanced lumen of the aorta from the rest of the data in the DICOM images.
WHY YOU ARE DOING IT: To identify the voxels that will eventually be represented in the 3D printed
model.
HOW TO DO IT: The two segmentation tools that will be used are “Thresholding” and “Region Growing”.
Both may be familiar from experience with standard 3D visualization. 1) Thresholding isolates voxels
with attenuation within a specified Hounsfield Unit (HU) range. 2) In region growing, the user manually
identifies a seed point and the software selects voxels within the specified HU range that are physically
connected to that seed point.
Each of these steps creates a “mask” or intermediate model that could be printed after further
manipulations. A list of the masks you have created appears in the first pane of the Project
Management Toolbar (this is located on the top right of your screen).
1. From the Segmentation Menu in the top Toolbar, choose Thresholding
. This opens a
window in which you can specify a HU range. This step creates a mask containing only the
pixels that fall within the specified HU range.
Set the HU range from 351 to 1399 to eliminate those tissues that fall outside 351-1399 HU.
Bone and the contrast material in the aorta will appear green in the images.
2. From the Segmentation Menu, choose Region Growing . This tool creates a new mask
(yellow) containing only those voxels within the source (green) mask that are connected to the
seed point that you identified.
3. Left-click on a point within the aorta (from any of the three planes) to specify a seed point.
This will highlight the aorta and its major branches in yellow, while the rest of the highlighted
tissue from the previous mask (e.g. bone) remains in green. Close the Region Growing tool.
4
The yellow arrow in the screen shot above indicates where you should left click (within the
aorta) to perform the region growing (step 3 above).
Hint: Make sure the Multiple Layer option box is selected. This will perform the operation throughout the entire
image stack.
4. Next, we will “Calculate” a 3D rendering of the aorta from the yellow mask. This
intermediate step allows you to visualize the result of the two segmentation steps combined.
From the Segmentation Menu, choose Calculate 3D . Ensure that the yellow mask is
highlighted and the Quality is set to Optimal and hit the Calculate button to create a “3D
object”.
A list of the 3D objects you have created appears in the second pane of the Project
Management toolbar (on the right, second from the top in the Project Management Toolbar).
To adjust the visualization, zoom with the mouse wheel and pan by holding the wheel down
and moving the mouse. To show the rendering on the full screen, either hover the mouse
cursor over or click the bottom right image and hit the spacebar. The screen can be reset to the
4-image view by hitting the spacebar again.
After viewing the 3D rendered volume, hide the object by clicking on the eyeglasses
Project Management Toolbar.
in the
5
Task B. Editing the Mask of the Aorta
WHAT YOU ARE DOING: Further segmentation of the DICOM images to remove most of the branch
vessels, followed by generation of an STL file (a different file format than DICOM that is recognized by
3D printers). The STL file then undergoes a smoothing step smoothing called “wrapping”.
WHY YOU ARE DOING IT: The product of each segmentation step we have performed thus far could
be converted into a STL format and sent to the 3D printer; however, we desire to simplify the model and
limit the amount of printing material needed by excluding the mesenteric branch vessels.
HOW TO DO IT: The first part of this task uses a third segmentation tool, “Edit Mask in 3D”, that
modifies the volume. This tool is not typically available in standard 3D visualization. We will use the
Lasso tool from this tool kit to select a small portion of each proximal branch vessel near the ostium,
which can then be “removed”. This breaks the voxel-to-voxel connection between the proximal and
distal portions of individual branches of the (e.g., the superior mesenteric artery). Then, after Region
Growing is reapplied, the distal branches are no longer rendered. The STL file will then be further
refined using a smoothing tool called “Wrap” that eliminates rough areas and gaps in the model.
Familiarize yourself with the following 3D interface shortcuts that will be used to edit your
model in 3D.
Shortcut
Right mouse button
Shift + Right mouse button
Center click on the mouse
Ctrl + Right mouse button
ArrowUp / PageUp
ArrowDown / Page Down
Arrow Right / End
Arrow Left / Home
Action
Rotate: Move the mouse to rotate
Pan: Move the mouse to pan
Zoom: Move the mouse vertically to
zoom
Rotate Up with discrete steps
Rotate Down with discrete steps
Rotate Right with discrete steps
Rotate Left with discrete steps
1. From the Segmentation Menu, choose Edit Mask in 3D
to edit the mask in the 3D
window. After choosing Edit Mask in 3D
, the bottom right image will be the yellow aorta
surrounded by a 3D transparent box, which indicates the fact that you are editing in 3D as
opposed to 2D. Hit the space bar to enlarge the view. Click the right mouse button and move
the mouse to rotate the image, and hold down the scroll wheel to pan the image so that the
aorta is viewed laterally with the mesenteric arteries pointing to the right of the screen.
2. Here is an image of the Edit Mask in 3D toolbar that appears at the top of the screen:
We will use the Lasso tool to select small portions of the mesenteric arteries. Draw a loop
around the portion you desire to remove. When the loop of the lasso has completely
surrounded the vessel, the encircled area to be removed will appear as a highlighted area in
6
the model. Lasso small proximal portions of the celiac, SMA, IMA, and lumbar branches, as
well as the mid-renal arteries.
3. Hit the Remove button to disconnect the proximal from distal arteries. This will remove
that portion of the artery, thereby disconnecting the distal mesenteric branches. Your lasso
edits need not be perfect. Close the Edit mask in 3D toolbar, and then hit the spacebar to
return from the enlarged view to the 4 window view.
Hint: The changes that you apply in this operation will only affect the mask information. It will not affect any 3D
models that you previously calculated.
4. From the Segmentation Menu, choose Region Growing . Left mouse click within the
lumen of the aorta (as you did earlier). This will remove the voxels that were disconnected in
the previous step. There is now a new mask called “Cyan”. Rename this mask, “Aorta” by
double-clicking on the name in the Masks tab of the Project Management Toolbar and typing
the word Aorta. Note that at this stage, the bottom right window is empty—we are going to
fill it by recalculating a new 3D model.
7
5. From the Segmentation Menu, choose Calculate 3D . You will see a window that
allows you to select a Mask and a Quality setting to create a 3D model. Choose your Aorta
mask and the Optimal setting. Press the Calculate button.
Note: This calculates a 3D surface model of the region that was highlighted in the “Aorta” mask. This model is
an STL file that describes the geometry as a set of connected triangles. STL is the file format needed to create a
3D printed geometry. This differs from a volume rendered model (i.e., 3D visualization) in that it contains
exportable surfaces.
6. Change the color of your model by selecting your Aorta model and clicking on the
Properties icon in the 3D Objects tab of the Project Management Toolbar. Click on ‘Artery’
in the ‘Type’ drop-down menu within the 3D Properties Window. Click ‘OK’.
7. From the Tools Menu, choose Wrap
to eliminate gaps and smooth rough areas on your
model. The Smallest Detail in the new window should be set at 0.5 mm; this is on the order
of the size of the CT voxel. This will eliminate rough areas representing image noise. The
8
Gap Closing Distance should be set at 2.0 mm. The gap closing distance refers to the largest
separation between points for which anatomic variations will be smoothed.
9
Task C. Exporting the Aorta model to 3-matic (CAD software)
WHAT YOU ARE DOING: Exporting the STL file from Mimics to 3-matic to perform final postprocessing steps. In 3-matic, the 3D rendering will undergo one level of smoothing. Then, the vessel
wall will be rendered around vessel lumen. Finally, we will label the model before the final STL file is
generated.
WHY YOU ARE DOING IT: To refine our 3D model to eliminate surface imperfections in the 3D printed
model. We will create a rendering of the aorta wall rather than the opacified intraluminal blood pool so
the printed model can be used for benchtop tasks such as evaluation of device deployment or
assessment of fluid flow dynamics †. The model will be labeled for identification purposes.
HOW TO DO IT: We will export the final iteration of our STL file from Mimics into 3-matic. In 3-matic, we
will perform an additional smoothing step. A rendering of the aorta wall will be generated using the
“Hollow” tool. We then will visualize the lumen within the hollow model after cutting the vessel endings
using the “Trim” tool. “Quick label” will be used to apply a customized label.
The screen shot below is displayed to familiarize you with the 3-matic interface.
1. From the 3-matic Menu in Mimics, choose Design . Select the “Wrapped_Aorta” model
and press, ‘OK’. This will open the model in the 3-matic software package.
It is important to recognize that all of the initial processing steps are performed on the opacified lumen
of the aorta, based on the fact that we initially segmented the high attenuation contrast material. Note
that it can be very challenging to segment and 3D print very thin structures such as the normal arterial
wall. Moreover, the intermediate attenuation of the actual wall (~40-60 HU) can limit the utility of tools
such as thresholding that act on the entire data set. One solution is to generate and print a
representation of the wall with a uniform thickness of 2 mm. This ‘artistic license’ can be used in 3D
printing. However, this simplification has important consequences—anatomic detail in the aorta,
including pathology in the wall, may be lost when the model is printed. Note that the representation of
the lumen remains accurate, based on the completed segmentation steps.
†
10
2. From the Finish Menu, choose Local Smoothing
. Set the ‘Smoothing diameter’
within the Operations Tab to 10 (millimeters). To smooth out remaining rough areas, zoom in
on the picture by using the mouse wheel. Then hold the left mouse button down over the
region to be smoothed while making small circular movements.
3. From the Design Menu, choose Hollow
to create the wall of the aorta. Copy the
parameters in the screen shot below, and then click ‘Apply’. It will take some time for this
operation to finish.
4. To simultaneously visualize the wall and lumen renderings, right click the Wrapped Aorta
model in the upper right window called “Scene Tree” and select Transparency > Medium.
This will demonstrate the wall as a shadowed area around the bright aorta lumen.
11
5. From the Finish Menu, choose Trim.
Note that at present, the ends of the model are
“closed” and the lumen cannot be seen. To open the ends of the model, free hand draw a box
(see screen shot below) around the proximal aspect of the aorta and press Apply. To
completely “open” the model, freehand draw additional boxes around the distal aspects of
both distal common iliac arteries. Press ‘Apply’ after you generate each box to cut each
respective portion.
12
The resulting model should now be cut open at the ends, enabling visualization of the lumen. ‡
‡
Prior to exporting the STL, it is possible to verify the accuracy of the completed 3D model
against the original DICOM data. Due to time constraints, this is covered in Appendix 2.
13
6. Rotate the model sidewise. We are now going to add a label. From the Finish Menu,
choose Quick Label
. You can label your 3D model with your name and “RSNA 2014”
in the “Text” box under Label parameters (screen shot below). Change the font height to
3.0000. Click on the model where you would like to place the label (if you press ‘Ctrl’ while
applying the label, it will be inset rather than raised off the surface).
The model will now bear a label with the information you entered.
7. From the Export Menu, choose Binary STL. Note that this is the step that generates the
STL file that will be exported to the 3D printer. Define the output directory as the desktop.
14
Task D. 3D printing the STL file
WHAT YOU ARE DOING: Using a new software platform called Objet Studio to place the aorta on the
“build tray”, the platform on which the model will be built. After visualizing the aorta from different
perspectives, we will assign the material and “send” the job.
WHY YOU ARE DOING IT: In Mimics and 3-matic, we have generated the STL file. Objet Studio is a
software package associated with the 3D printer used to organize print jobs, select the materials, and
execute printing.
HOW TO DO IT: The STL file will be imported into the new software package. Next, the file is rendered
as it will be printed on the build tray. The orientation can then be adjusted, and the material selected.
We have now switched software packages and will simulate the printing of the aorta model
using Objet Studio.
1. Open Objet Studio by double-clicking on the Objet Studio icon:
A dialog box will open, asking you to select a printing server connection. We do not have a
server at RSNA, so please use “Unknown”. This will allow us to work off-line.
The main screen of Objet Studio will open as shown in the screen capture below:
15
2. To import the aorta STL file, click the upper left menu item “Insert”, and select your
model from the dialog box that opens:
Your model will be automatically imported onto the software representation of the build tray
of the printer. The build tray is shown in blue below. The orientation of the aorta is optimized
so that the support material needed for the 3D printing job can be minimized (It would be
possible, but far less efficient, to print the aorta standing “upright” rather than “lying down”).
3. Model visualization. As shown in the prior screen capture, the default view is of the entire
build tray of the 3D printer in a single isometric view. In the tab “Model Settings” you can
zoom into the model in this view by clicking the icon “Zoom Tray”, and selecting “Zoom
Selected Object”. The result will appear as follows:
16
In practice, with several objects on the build tray, it is important to be able to change the
orientation of the model. This can be done by choosing “Isometric View”, and navigating the
pull-down menu to “Northeast Isometric”, then repeating the zoom by clicking the icon
“Zoom Tray” and selecting “Zoom Selected Object”. This series of key stokes will move
the model on the tray with the result as follows:
For some jobs it will be advantageous to study the model in a 4-axis view. To do this, click
the icon “Single View” and select “4 View”. To make your screen appear as below, click on
each individual window, make sure that the abdominal aorta is selected by single-clicking on
it, and select “Zoom Object” for a close-up view.
17
To complete further operations on this model, return to the single window view by clicking
the “4 View” icon and then select “Single View”. Then click “Isometric View”, and select
the “Northeast Orientation” in the pull-down menu. Zoom into the selected model by
clicking “Zoom Object”. You should have on your screen the following:
4. Next, we will assign the material to be used to build the model and start the 3D printing job.
Click on the left-most tab labeled “Tray Settings”. We will change the material settings to
Tango+, an elastomeric material. To do this, select the first material pull-down menu on the
right of the “Tray Setting” menu that says “VeroBlack”, and select “Tango+”. Your model
will change color, reflecting the printing in Tango+. There will be a confirmatory dialog box
pop-up.
In a production environment where the printer is connected to a print server, you would then
proceed to the “Job Manager” menu tab to connect to the printer, send the job, and ensure
that the material store in the printer is sufficient to complete the job.
18
APPENDIX 1. Importing DICOM images into Mimics.
1.
Open Mimics 17.0 from the Start Menu or desktop icon.
2.
Select the New Project Wizard from the File Menu.
3.
Browse to the location of your DICOM images. Click Next.
4.
The following window will show the studies contained within the folder. This gives a preview of
the images and the DICOM tags. Choose Convert.
5.
Indicate the orientation in the orientation window. After clicking ‘OK’ the Mimics project is
opened and work can be started.
APPENDIX 2. Verifying the Accuracy of the Model
Before exporting the model for 3D printing or using it for any further design, you may want to check that the
model is an accurate representation of the patient’s anatomy by bringing it back into Mimics and comparing it to
the original images.
1.
In 3-matic, click on the model name in the Scene Tree. Copy it by using CTRL+C on your
keyboard.
2.
Go back to Mimics project and paste the model using CTRL+V. You will see it appear in the STLs
tab of the Project Management Toolbar.
3.
Make the contours of the Aorta model visible in the image views by clicking on the ‘Contours
Visible’
4.
sunglasses icon in the STLs tab.
Scroll through the images to ensure that the model accurately follows the geometry of the rendered
anatomic structure.
19
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement